plan: #3133

[PolyphonyRequiem]

---

requests_parent_change: false
ancestor_plan_generations:
root: 1

Plan for #3133 (root #3131)

Promotes plan/3131-3133 into plan/3131.

AB#3133

[PolyphonyRequiem]

🤖 Polyphony plan reviewer (advisory)

Summary: A tight, well-scoped first-pass slice that correctly identifies the research_needs schema as the binding contract for downstream Issue #3136 and sequences the three tasks accordingly. Scope is realistic for a single PG, archive-only constraint is explicit, and harness coverage is built in. Main gaps are around schema location/shape ambiguity, missing detail on the findings-document shape (which is itself a contract), and acceptance criteria that lean qualitative in a couple of spots.

Strengths

• Correctly frames the schema as the stable contract that #3136 binds to, and sequences task-1 first for that reason.
• Archive-only / cheap-tier scope is unambiguous and explicitly excludes live tools — good guardrail against scope creep into #3134.
• Harness-driven end-to-end scenario with a scripted FakeProvider is built into AC, not deferred.
• References the right skills (polyphony-workflow-author, polyphony-harness, conductor-mechanics) and the canonical config paths.
• Dependency call-out (downstream #3136, #3134) is accurate and helps the reviewer of those issues.

Concerns

• Schema location is left "TBD by the workflow-author skill." Since this schema is the binding contract for #3136, leaving its location unresolved at plan time risks task-1 stalling on a placement decision, or #3136 binding to a path that later moves. Resolve before implementation starts.
• Findings-document shape is under-specified. The plan says "a simple, documented shape" but the document is itself a contract consumed by the architect leg in #3136. Without at least a sketch of required sections/frontmatter, task-3's harness assertion ("asserts the findings document shape") is not yet testable.
• budget_hint: cheap | extended in the schema implies awareness of an extended tier that doesn't exist in this slice (#3134 territory). Either drop it from this pass or document that it is forward-compatible scaffolding so reviewers don't expect routing behavior here.
• AC #3 ("sources strictly from the archive") is not directly verifiable by the harness scenario as described — the FakeProvider is scripted, so absence of live tool calls is by construction. Consider adding a structural check (e.g., agent-guidance / tool-allowlist assertion) so the constraint is enforced beyond this pass.
• No explicit mention of state-name handling. Not a blocker for a sub-workflow that doesn't transition work items, but worth a one-line confirmation that no hard-coded state strings are introduced (recent regressions: 9f96f8b, 03aab89, 5ea9929; canonical pattern scripts/scope-closer.ps1:54-60).

Suggested changes

• Pin the schema path in this plan (e.g., .polyphony-config/schemas/research-needs.schema.json) rather than deferring; #3136 needs a stable target.
• Add a short "Findings document shape" subsection (required headings, any frontmatter) so task-3's AC is testable and #3136 has something concrete to consume.
• Either remove budget_hint from the task-1 schema or label it explicitly as forward-compat for #3134.
• Strengthen AC #3 with a structural assertion (tool-allowlist or agent-guidance check), not just behavioral absence under a fake provider.
• Add an explicit "no hard-coded state-name strings introduced" line to AC as a standing guardrail.

This is an advisory comment from the polyphony plan_reviewer agent. It does not vote — human review is the merge gate.

[PolyphonyRequiem]

polyphony:approve 5463f48